«Published: October, 2010 qlikview.com Table of Contents An Introduction to the QlikView Associative Experience 3 Queries and Cubes Create a Blind ...»
THE ASSOCIATIVE EXPERIENCE:
QLIKVIEW’S OVERWHELMING ADVANTAGE
A QlikView Technology White Paper
Published: October, 2010
Table of Contents
An Introduction to the QlikView Associative Experience 3
Queries and Cubes Create a Blind Spot 3
MOLAP, ROLAP, and HOLAP: Maybe Fast, Maybe Flexible, But Not Associative 3 QlikView’s Secret: Our Associative Architecture 5 By Maintaining Associations, QlikView Helps Answer Un-Asked Questions 7 A Few Analogies: Maps, Engines, and the Rubik’s Cube© 8 QlikView is Quick to Deploy, Easy to Use, and Rich with Insights 10 People Love QlikView and There’s a Technical Reason for It 10 Appendix 11 The Associative Experience | Page 2 An Introduction to the QlikView Associative Experience This QlikView Technology White Paper illuminates the inner workings of QlikView’s associative architecture and explains how it fundamentally improves the way people conduct business analysis. The core question this paper answers is, “What is an ‘associative experience’ and how does it help people make the best business decisions?” To do so, we will explain how fundamental differences in underlying technology make QlikView’s associative experience different from the experience people have with other BI tools. Our target audience consists of those business decision makers who are evaluating business intelligence (BI) software, and the technology decision makers supporting them.
Queries and Cubes Create a Blind Spot By their nature, most query-based tools divorce data from their context, leaving gaps for those trying to make data-driven business decisions. Those making complex business decisions don’t always have full access to their supporting data — even when they have access to BI software. Some data is available only as isolated and discrete queries, without context between one query and the next.
Three drawbacks to traditional query- and cube-based BI tools are:
• They leave data on the table. Query-based tools extract a small subset of data from the main dataset. They aggregate extracted data and return it in the form of a query result set. This result set is completely divorced from any other data not contained in the individual query. The very act of extracting the subset of data from the main dataset breaks associations.
• They create a dissociative experience. With cube-based technology, each individual query represents a single discrete chunk of information. This information is extracted from the underlying database or cube and passed on to the user as a discrete set of data. If the user wants to know more about how a piece of data contained in a query relates to another piece of data outside the query, the only choice is to formulate a new query that incorporates both data points and resubmit it.
• They don’t maintain relationships among queries. While the least sophisticated query-based tools don’t even attempt to provide context by tracking relationships among queries, some of the more sophisticated tools at least try. But this is a huge task. Each query is different, and most are not well-defined until users begin their analyses. So if it’s not possible to know what query a user will formulate next, how can we provide associations between queries?
MOLAP, ROLAP, AND HOLAP: MAYBE FAST, MAYBE FLEXIBLE,
BUT NOT ASSOCIATIVEOnline analytical processing (OLAP) uses aggregated data for decision support. Many variations of OLAP exist. Some are flexible and others are high-performance. But as long as they are query-based, they are unable to maintain associations among data elements (see Figure 1).
Source: QlikTech, Inc.
For the past 50 years, query-based architectures have been the status quo for decision support. The ubiquity of structured query language (SQL) creates a blind spot to the shortcomings of using queries — whether SQL, multidimensional query expressions, or otherwise — as the fundamental component of a decision support engine.
• ROLAP extracts data in real-time as it is needed, making it flexible. The oldest form of OLAP decision support is relational online analytical processing (ROLAP).
ROLAP is still prevalent today. It uses SQL or other query technology to extract and calculate data aggregates in real time as the user needs them. Once thought of as slow and unresponsive, today ROLAP is enjoying something of a renaissance with the more scalable decision support database architectures. ROLAP can be flexible, without requiring predefined dimensionality, but is computationally intensive and can therefore be slow. And because ROLAP is query-based, it is unable to maintain associations.
• MOLAP pre-aggregates data, making it fast. The next generation of technology for decision support came in the form of multidimensional online analytical processing (MOLAP), also known as cube-based OLAP. The main difference between ROLAP and MOLAP is that with MOLAP the query results are aggregated in advance while for ROLAP they are aggregated as needed. With MOLAP, data is pre-aggregated for multiple permutations of data points along preselected dimensions. This approach provides nearinstantaneous access to aggregates as long as the question the business user has in mind lies within the predefined dimensionality. Because the aggregates are pre-calculated, MOLAP can be faster than ROLAP. However, with this speed comes a loss of flexibility.
And again, because MOLAP is query-based it cannot maintain associations.
• HOLAP offsets some ROLAP and MOLAP weaknesses. The relative strengths and weaknesses of ROLAP and MOLAP led to the creation of a third technology: hybrid online analytical processing (HOLAP). HOLAP is any architecture that leverages both ROLAP and MOLAP in an attempt to offset the relative weaknesses of each. Because HOLAP is the product of the marriage of two query-based technologies, it is also a fundamentally a query-based technology. And — you guessed it — it does not maintain associations in the data.
In contrast, QlikView is flexible, fast, and maintains associations among all data elements.
QlikView offers the flexibility of ROLAP (no predefined dimensionality) as well as the speed of MOLAP (near-instantaneous access to aggregates). While MOLAP tools sometimes have drill-through capabilities (in essence, a multidimensional engine with on-demand relational queries), QlikView is just the opposite: a relational engine with on-demand cubes.
The Associative Experience | Page 4 QlikView’s Secret: Our Associative Architecture QlikView delivers the world’s first associative architecture (see Figure 2 and Figure 3).
QlikView manages associations among data sets at the engine level, not the application level. QlikView stores individual tables in its in-memory associative engine. Every data point in every field is associated with every other data point anywhere in the entire schema.
Datasets can be hundreds of tables with thousands of fields.
Figure 2: Qlikview’s underlying architecture is associative
Source: QlikTech, Inc.
Query-based BI tools separate the application layer from the data layer. This leads to long deployments while expensive developers customize the application layer to manage the specific associations required to answer a particular business question. When the BI application needs to answer a slightly different business question, the application layer must be altered again. This process is time-consuming and expensive.
The Associative Experience | Page 7
A FEW ANALOGIES: MAPS, ENGINES, AND THE RUBIK’S CUBE®This is complicated stuff. How about an analogy or two?
First, let’s say the goal is to understand how an internal combustion engine works using digital models. With the query-based paradigm, we would look at individual parts of the engine in isolation (see Figure 5). We would see one part at a time. We would be left to attempt to understand the relationships (or associations) among the parts and how the parts all fit together as a cohesive whole to create a working engine.
Figure 5: Analogy #1 — Understanding an internal combustion engine
+ + Source: QlikTech, Inc.
Using QlikView’s associative technology, however, we now have access to a digital model of a complete working engine with each part in correct relationship to all of the other parts. We can tweak the throttle in the digital model (or execute a selection, in QlikView) and see how that affects the fuel intake, carburetor, and exhaust. We can watch the pistons pump and turn the crankshaft. We can deconstruct the engine at our leisure and look at each part in context with the parts next to it. This is the power of QlikView’s associative architecture.
In another analogy, imagine a Rubik’s Cube puzzle — but to solve it you can only see one face at a time (see Figure 6). As you change one face, you cannot see what is going on with the other sides. In contrast, using QlikView is like being able to see all faces of the Rubik’s Cube at the same time, so you can understand what else is changing based on changes you are making.
Source: QlikTech, Inc.
In a third example, let’s say you are planning a trip from London to Rome. A traditional road atlas can help. But the atlas is time-consuming because you have to relate how the map fits together across multiple pages in the atlas. In addition, you may miss details as you are consumed by following a particular road. QlikView is more like Google® maps: you can see the entire route at once or zoom in on areas of special interest. You can identify better routes quickly based on traffic patterns. Importantly, you become far more engaged with the interactive map than with the atlas.
The Associative Experience | Page 9
QLIKVIEW IS QUICK TO DEPLOY, EASY TO USE, AND READY TO
QlikView customers love us. Here’s why:
• QlikView is quick to deploy. Because QlikView’s associative architecture does not need to be reconfigured to answer new business questions, QlikView is quick to deploy.
According to the IT team at FHL Banks, “Without QlikView, it would take a year and a million dollars to implement this functionality with traditional BI.” Radiometer told us that “QlikView enables us to analyze data much faster and to react much faster. Generating reports only take 25% of the time that used to be needed.” Agora Publishing deployed QlikView in 20% of the forecasted time and came in 60% under budget. And the CIO of Superior Graphite Company said, “In four hours, one person achieved with QlikView what teams of IT consultants … couldn’t achieve in a year and a half.”
• QlikView is easy to use. Because QlikView’s associative architecture preserves all of the associations between all fields being analyzed, users find it intuitive and easy to use. According to the IT director at Bliss, “Users love QlikView and how they can instantly access data for planning new campaigns or running inventive sales promotions. With QlikView, employees are now more accountable for performance and truly empowered with the visibility to make the best business decisions possible at that moment in time.” An analytics project manager at a global investment bank said, “QlikView is so flexible, powerful and easy to use—views of data have not been seen this way previously. Our salespeople are winning more business from clients directly as a result of QlikView.”